Tuesday, May 14, 2013

The number one question Hillary's pollsters need to ask

It's spring in an odd-numbered year, a quiet moment in national electoral politics. Political junkies pay too much attention to special elections, and then write about what they mean, and what they don't mean, and how we shouldn't draw any conclusions from them anyway.

It's also a moment when issue groups, taking the long view of electoral politics (at least in comparison to individual candidate campaigns) begin to set the stage for the following year and the next Administration. The big news on that front this month was the Madam President initiative launched by Emily's List. It's meant to be a full-throttle push for "any" woman president, but everyone knows that in 2016 that really means Hillary Clinton.



When she runs (and I believe she will run) she will certainly have a few things to learn from the Obama campaigns in 2008 and 2012. Much fanfare has been given to the Obama use of technology and micro-targeting techniques, but it may be several decades before analysts fully unpack the dynamics of those campaigns. One book attempting to do so is The Victory Lab by Sasha Issenberg. It's an extremely well-researched deep dive into the context and mechanics of how the Obama campaigns leveraged computer technology, statistics and polling to win the day. The overused but apt comparison is Moneyball for politics. Secretary Clinton's staff will absorb these techniques and build upon them for sure.

But here's where I think it gets interesting for women: statistical models that make predictions of voting behavior based on demographic and consumer information require past behavior to create the algorithms. Whenever you have a "first" candidate, the predictive power of the model is automatically weaker. Issenberg notes that the Obama campaign in 2008 made specific changes to its polling practices and resulting predictive models to avoid what is known as the Bradley Effect, where a voter will tell a pollster s/he will vote for a minority candidate (or is undecided) when in fact the voter ends up voting for the white candidate because of race. The candidate can appear ahead in the polls but still come up short at the polling place.

As a strategy, the Obama campaign didn't seek to change the minds of these voters. Instead, they simply wanted to exclude them from the get-out-the-vote efforts, which were the heart of the campaign's strategy. Identifying these voters turned out to be remarkably simple. According to Issenberg, the pollsters began to ask, "Do you think your neighbors would be willing to vote for an African-American president?" It turned out that most of the time, behavior attributed to "the neighbors" was really the voter's own, even if they couldn't admit it.

The numbers on likelihood to vote for a woman are surely too astonishing right now: 90% say they would, and 72% say they believe it's likely a woman will win in 2016. Could it be that simple to straighten this out? Is all we need to ask, "Would your neighbors vote for a woman president?" to get a clearer picture?

I promise this is not a test, but would your neighbors vote for a woman president? Would they vote for Hillary?

No comments:

Post a Comment